Subtype Classification of Renal Parenchymal Tumors on MLP-Based Methods

Renal parenchymal tumors are among the most common tumors in humans. With the development of deep learning, it has become possible to use deep neural networks to distinguish renal parenchymal tumor subtypes. This paper aims to investigate the role of the Multilayer Perceptron (MLP) structure in the classification of renal parenchymal tumor subtypes on magnetic resonance (MR) images. We design a classification model based on ConvMLP. In addition, we introduce Convolutional Block Attention Modules (CBAMs) on the basis of ConvMLP to further improve the classification precision. In order to find where adding CBAMs improves the performance the most, we design four variant networks. We conduct extensive comparative experiments on these four variant networks and other convolutional neural networks. The experimental results show that the addition of CBAM improves the classification precision of renal parenchymal tumor subtypes by 3%, and compared with other CNNs, our classifier has the highest precision.